For Devs


The official form of an ISBN is something like ISBN 979-10-90636-07-1. However for most applications only the numbers are important and you can always masked them if you need (see below). This library works mainly with ‘striped’ ISBNs (only numbers and X) like ‘0826497527’. You can strip an ISBN’s like string by using canonical(isbnlike). You can ‘mask’ the ISBN by using mask(isbn). So in the examples below, when you see ‘isbn’ in the argument, it is a ‘striped’ ISBN, when the argument is an ‘isbnlike’ it is a string like ISBN 979-10-90636-07-1 or even something dirty like asdf 979-10-90636-07-1 bla bla.

Two important concepts: valid ISBN should be an ISBN that was built according with the rules, this is distinct from issued ISBN that is an ISBN that was already issued to a publisher (this is the usage of the libraries and most of the web services). However, probably by legal reasons, merges the two! So, according to, ‘9786610326266’ is not valid (because the block 978-66… has not been issued yet, however if you use is_isbn13('9786610326266') you will get True (because ‘9786610326266’ follows the rules of an ISBN). But the situation is even murkier, try meta('9786610326266') and you will see that this ISBN was already used!

If possible, work with ISBNs in the isbn-13 format (since 2007, only are issued ISBNs in the isbn-13 format). You can always convert isbn-10 to isbn-13, but not the reverse. Read more about ISBN at

API’s Main Namespaces

In the namespace isbnlib you have access to the core methods:

Validates as ISBN-10.
Validates as ISBN-13.
Transforms an isbn-13 to isbn-10.
Transforms an isbn-10 to isbn-13.
Keeps only numbers and X. You will get strings like 9780321534965.
Cleans ISBN (only legal characters).
notisbn(isbnlike, level='strict')
Check with the goal to invalidate isbn-like.
get_isbnlike(text, level='normal')
Extracts all substrings that seem like ISBNs (very useful for scraping).
get_canonical_isbn(isbnlike, output='bouth')
Extracts ISBNs and transform them to the canonical form.
Transforms an isbnlike string into an EAN13 number (validated canonical ISBN-13).
Gets the language or country assigned to this ISBN.
mask(isbn, separator='-')
Mask (hyphenate) a canonical ISBN.
meta(isbn, service='default')
Gives you the main metadata associated with the ISBN. As service parameter you can use: 'goob' uses the Google Books service (no key is needed) and is the default option, 'wiki' uses the api (no key is needed), 'openl' uses the api (no key is needed). You can enter API keys with config.add_apikey(service, apikey) (see example below). The output can be formatted as bibtex, csl (CSL-JSON), msword, endnote, refworks, opf or json (BibJSON) bibliographic formats with isbnlib.registry.bibformatters.
editions(isbn, service='merge')
Returns the list of ISBNs of editions related with this ISBN. By default uses ‘merge’ (merges ‘openl’, ‘thingl’ and ‘wiki’), but other providers are available: ‘openl’ uses Open Library, ‘thingl’ (uses the service ThingISBN from LibraryThing), ‘wiki’ (uses the service Citation from Wikipedia) and ‘any’ (first tries ‘wiki’, if no data, then ‘openl’ or ‘thingl’).
Returns the most probable ISBN from a list of words (for your geographic area).
Returns a list of references from Google Books multiple references.
Returns a dictionary of classifiers for a canonical ISBN. For the meaning of these classifiers see OCLC. Most of the data in the underlying service are for books in english.
Returns a DOI’s ISBN-A from a ISBN-13.
Returns metadata formatted as BibTeX for a given DOI.
Renames a file using metadata from an ISBN in his filename.
Returns a small description of the book. Almost all data available are for US books!
Returns a dictionary with the url for cover. Almost all data available are for US books!

See files test_core and test_ext for a lot of examples.

The exceptions raised by these methods can all be caught using ISBNLibException.

You can extend the lib by using the classes and functions exposed in namespace, namely:

  • WEBService a class that handles the access to web services (just by passing an url) and supports gzip. You can subclass it to extend the functionality… but probably you don’t need to use it! It is used in the next class.
  • WEBQuery a class that uses WEBService to retrieve and parse data from a web service. You can build a new provider of metadata by subclassing this class. His main methods allow passing custom functions (handlers) that specialize them to specific needs (data_checker and parser). It implements a throttling mechanism with a default rate of one call per second per service.
  • Metadata a class that structures, cleans and ‘validates’ records of metadata. His method merge allows to implement a simple merging procedure for records from different sources. The main features can be implemented by a call to stdmeta function!
  • vias exposes several functions to put calls to services, just by passing the name and a pointer to the service’s query function. vias.parallel allows to put threaded calls. You can use vias.serial to make serial calls and vias.multi to use several cores. The default is vias.serial, but you can change that in the conf file.

The exceptions raised by these methods can all be caught using ISBNLibDevException. You shouldn’t raise this exception in your code, only raise the specific exceptions exposed in whose name end in Error.

In you can find several methods, that we found very useful, some of then are only used in isbntools (an app and framework that uses isbnlib).

With isbnlib.config you can read and set configuration options: change timeouts with seturlopentimeout and setthreadstimeout, access api keys with apikeys and add new one with add_apikey, access and set generic and user-defined options with options.get('OPTION1') and set_option.

Finally, from isbnlib.registry you can change the metadata service to be used by default (setdefaultservice), add a new service (add_service), access bibliographic formatters for metadata (bibformatters), set the default formatter (setdefaultbibformatter), add new formatters (add_bibformatter) and set a new cache (set_cache) (e.g. to switch off the cache set_cache(None)). The cache only works for calls through metadata functions. These changes only work for the ‘current session’, so should be done always before calling other methods.

Let us concretize these points with a small example.

Suppose you want a small script to get metadata using Open Library formatted in BibTeX.

A minimal script would be:

from isbnlib import meta
from isbnlib.registry import bibformatters

SERVICE = "openl"

# now you can use the service
isbn = "9780446310789"
bibtex = bibformatters["bibtex"]
print(bibtex(meta(isbn, SERVICE)))

All these classes follow a simple design pattern and, if you follow it, will be very easy to integrate your classes with the rest of the lib.


You can extend the functionality of the library by adding pluggins (for now, just new metadata providers or new bibliographic formatters).

Start with this template and follow the instructions there. For inspiration take a look at goob.

After install, your pluggin will blend transparently in isbnlib.

Remember that plugins must support python 2.7 and python 3.5+ (see

For available pluggins check here.

Extra Functionality

To get extra functionality, search pypi for packages starting with isbnlib or type at a terminal:

$ pip search isbnlib

for a nice formatted report!

Merge Metadata

The original quality of metadata, at the several services, is not very good! If you need high quality metadata in your app, the only solution is to use polling & merge of several providers and a lot of cleaning and standardization for fields like Authors and Publisher.

You can write your own merging scheme by creating a new provider.


These classes are optimized for one-calls to services and not for batch calls.